From Protein Structures to Peptide Therapeutics

From Protein Structures to Peptide Therapeutics: A 5-Phase Pipeline

Protein structure to peptide therapeutic development follows a five-phase pipeline: Target Characterisation, Lead Identification, Lead Optimisation, Preclinical Validation, and Clinical Translation. In each phase, deposited structural-genomics coordinates drive the critical decision. The GLP-1 receptor series illustrates this progression concretely: semaglutide and tirzepatide, both now EMA-approved for type 2 diabetes and weight management, were shaped by cryo-EM datasets that resolved receptor-peptide contacts at atomic resolution.

This article covers how deposited protein structures constrain each phase of peptide drug discovery; why multi-conformation datasets outperform single-snapshot structures for lead optimisation; and how the GLP-1 agonist series (exenatide, liraglutide, semaglutide, tirzepatide) demonstrates this pipeline in practice, from natural peptide scaffold to EMA-approved dual agonist.

The five phases, in order, are:

1. Target Characterisation, defining binding-site geometry from deposited coordinates
2. Lead Identification, screening peptide libraries against the structural model
3. Lead Optimisation, iterative co-crystal or cryo-EM feedback to improve potency and selectivity
4. Preclinical Validation, confirming mechanism in cellular and animal models using structure-informed analogues
5. Clinical Translation, carrying structural evidence into regulatory dossiers and dose-optimisation studies

Peptides are conventionally defined as chains of fewer than 50 amino acids, a working boundary that holds across structural-biology and medicinal-chemistry literature into 2026. Each phase below maps exactly which structural data are needed, how multi-conformation datasets change the decision, and where the GLP-1 series provides a dated, phase-annotated example.

Why a Deposited Structure Is the Starting Line, Not the Finish

A deposited protein structure is a decision-making tool: it encodes binding-site geometry, pocket volume, and conformational flexibility in a form that every downstream phase of protein structure to peptide therapeutic development can query, measure against, and build upon. The coordinate file sitting in the PDB is not the scientific conclusion. It is the evidence base from which lead identification, optimisation, and preclinical design all draw their first hypotheses.

Structural-genomics programmes formalised this logic at scale. The Joint Center for Structural Genomics (JCSG) deposited hundreds of crystal structures into the PDB with the explicit mandate of populating target space ahead of therapeutic need, treating deposition as a public-good act rather than a proprietary endpoint. The JCSG pipeline methodology, from gene cloning through crystallisation to coordinate release, was designed so that medicinal chemists and structural biologists could access coordinates without waiting for a disease-specific programme to fund the upstream work. That institutional model underpins the argument made throughout this article: publicly available coordinates compress the earliest phases of drug discovery by removing the need to solve a target structure de novo for every project.

The peptide/protein boundary matters here because it defines the chemical space the pipeline addresses. A working convention of fewer than 50 amino acids for peptides is consistent across structural-biology and medicinal-chemistry literature into 2026, though some authors apply physicochemical criteria (folding behaviour, molecular weight, or receptor-binding mode) rather than a strict residue count. For editorial consistency, the <50-residue heuristic is used throughout, with the acknowledgement that 30–40-residue hormones such as GLP-1 itself sit comfortably within it.

The five-phase pipeline proceeds as follows:

1. Target Characterisation, deposited coordinates define binding-site geometry and identify druggable pockets.
2. Lead Identification, structural models guide peptide-library screening and computational docking.
3. Lead Optimisation, iterative co-crystal or cryo-EM feedback refines potency, selectivity, and metabolic stability.
4. Preclinical Validation, structure-informed analogues are tested in cellular and animal models to confirm mechanism.
5. Clinical Translation, structural evidence enters regulatory dossiers and informs dose-optimisation studies.

Each phase below specifies which structural data are required, how multi-conformation datasets alter the decision, and where the GLP-1 series (including EMA-approved semaglutide and tirzepatide ) provides a dated, phase-annotated example.

Phase 1, Target Characterisation: Reading the Coordinates

Deposited structural-genomics coordinates define a drug target at the molecular level by revealing domain boundaries, conserved residue networks, surface topology, and the geometry of candidate binding sites before any compound is synthesised. This front-loaded structural intelligence is what separates a well-characterised target from a sequence entry with an assumed function.

What the Coordinates Actually Reveal

A single deposited structure provides a snapshot: chain termini that define stable domain boundaries, secondary-structure elements that constrain construct design for expression and assay development, and solvent-exposed pockets that score as druggable by cavity-detection algorithms. Domain boundary data alone has practical consequences. A construct that includes a disordered linker will aggregate in solution, whereas one trimmed to the folded core (as identified from the deposited model) will not. Structural-genomics pipelines formalise this step; the JCSG pipeline methodology documents how target selection, construct engineering, and crystallisation screening are treated as an integrated sequence rather than sequential hand-offs between separate groups.

The Unique Value of Multi-Conformation Datasets

Where structural genomics diverges from conventional single-structure determination is in the deliberate capture of the same target under varied crystallisation conditions, pH values, and ligand-free or ligand-soaked states. The ensemble that results exposes conformational heterogeneity that a single closed-state structure conceals. Cryptic binding sites, pockets absent in the dominant conformation but open transiently in minor populations, are a direct product of this approach. Optimising a lead against a single closed-state structure risks investing synthetic effort in a conformation the receptor rarely adopts in solution; multi-conformation datasets reduce that risk by showing which pockets are geometrically stable across multiple crystal forms. The JCSG Structure Gallery illustrates the institutional scale at which this multi-condition screening was applied across hundreds of targets during the Protein Structure Initiative era.

GLP-1 Case Study: Phase 1 Annotation

The GLP-1 receptor (GLP-1R) is a class B1 G protein-coupled receptor whose extracellular domain (ECD) was resolved crystallographically in multiple conformations before full-length receptor structures became accessible by cryo-EM. Those ECD structures defined the helical binding groove (a hydrophobic cleft that accommodates the N-terminal helix of GLP-1(7–36) amide) and established that the groove geometry is conserved across conformational states, making it a reliable pharmacophore anchor. Every subsequent GLP-1 agonist in clinical use, including semaglutide and tirzepatide, targets this same groove. The multi-conformation ECD data also revealed that the groove tolerates fatty-acid conjugation at specific positions without steric clash, a structural observation that directly informed the C-18 diacid linker strategy used in semaglutide's design. A single closed-state snapshot would have missed that tolerance entirely.

A structurally characterised target is one whose binding-site geometry is known across multiple conformational states, not merely in one crystal form. That distinction is the foundation on which every downstream phase of the pipeline is built.

Phase 2, Lead Identification: Screening Against the Structure

The binding-site geometry established in Phase 1 directly constrains the chemical space searched in Phase 2. Every lead-identification route, natural ligand mining, phage-display library screening, or computational peptide design, uses the deposited coordinates as a filter, not merely as background context.

Natural Ligand Mining

The most direct route to a lead peptide is to ask which endogenous or exogenous sequences already bind the characterised pocket. For the GLP-1 receptor, that question was answered by exendin-4, a 39-residue peptide isolated from the venom of the Gila monster (Heloderma suspectum). Exendin-4 shares 53% sequence identity with GLP-1(7–36) amide but carries a unique C-terminal tryptophan cage that stabilises its helical conformation in solution. Early receptor-bound structures showed that this helical C-terminal region docks into the extracellular domain groove identified in Phase 1, providing a pre-organised pharmacophore that GLP-1 itself (which is intrinsically disordered in solution) cannot reliably adopt. That structural observation made exendin-4 the template from which exenatide was developed, and the helical scaffold it defined has propagated through every subsequent GLP-1 agonist lead.

Phage-Display Library Screening

Where no natural ligand exists, phage-display libraries allow billions of peptide sequences to be screened against an immobilised or solubilised target. The deposited structure informs which positions must be held constant (residues that make direct hydrogen-bond or hydrophobic contacts with the binding-site wall) and which positions can be diversified. This is not a generic combinatorial search; it is a structure-directed one. Residues predicted by the Phase 1 coordinates to occupy the core of the pocket are typically fixed in the library design, while solvent-exposed positions are randomised. The JCSG pipeline methodology documents how structural data from high-throughput deposition campaigns were used in precisely this way to prioritise contact residues before library construction.

Library screening also surfaces structurally related non-peptide scaffolds. Peptoids (N-substituted glycine oligomers) share backbone geometry with peptides and can be synthesised via solid-phase methods and discovered through the same library formats. Fetse et al. (2023) demonstrated that peptoids retain structural motifs sufficient to engage peptide-binding pockets, making them relevant leads when proteolytic stability is a concern at this stage.

Computational Peptide Design

Structure-based docking and generative modelling against deposited coordinates represent the third route. Docking protocols place candidate sequences into the Phase 1 binding site and score them by predicted complementarity; generative models, including diffusion-based architectures, propose novel sequences conditioned on pocket geometry. AlphaFold 3-class models extend this to protein–peptide complex prediction, though as of 2026 their outputs are best treated as structural hypotheses requiring experimental validation rather than as primary evidence. The JCSG Structure Gallery illustrates the breadth of deposited coordinates available as docking targets across diverse receptor families, and the protein–peptide interaction resources at JCSG provide curated geometric parameters that inform scoring function calibration.

Across all three routes, the Phase 1 multi-conformation dataset is the operative constraint. A lead identified against a single closed-state structure risks failing when the receptor adopts an alternative conformation; leads validated against an ensemble of deposited states carry a structurally grounded confidence that single-snapshot screening cannot provide. Exenatide is a textbook illustration: its helical scaffold was geometrically compatible with the ECD groove across multiple crystal forms, a compatibility that single-conformation docking would have confirmed only partially.

Phase 3, Lead Optimisation: Co-Crystal Structures Drive Chemistry

Co-crystal structures of a peptide lead bound to its target receptor are the primary decision-making tool in Phase 3, converting a promising binder into a chemically stable, pharmacokinetically viable candidate through an iterative cycle of structure determination, synthetic modification, and re-determination against the modified compound.

The ordered optimisation workflow typically proceeds as follows:

1. Solve a co-crystal or cryo-EM structure of the Phase 2 lead in complex with the target.
2. Identify contacts lost to conformational flexibility, solvent-exposed backbone amide bonds susceptible to proteolysis, and regions tolerant of conjugation.
3. Apply a targeted chemical modification (stapling, N-methylation, or lipid conjugation) guided by the structural map.
4. Re-solve the complex to confirm that the modification preserves or improves binding geometry.
5. Advance only those variants whose structural re-determination validates the intended change; discard those that perturb the binding mode.

This cycle is not linear in practice. A stapled analogue that improves helical stability may shift a side-chain contact, requiring a further round of backbone adjustment before the pharmacokinetic modification is introduced.

Peptide Stapling

Hydrocarbon or lactam bridges that covalently constrain adjacent turns of an alpha-helix reduce the conformational entropy penalty on binding and simultaneously shield backbone amide bonds from endoproteases. Constrained and cyclic peptides are described in the formulation literature as therapeutic protein mimetics locked in specific conformations to ensure therapeutic efficacy. Hayward et al. (RSC, 2024) document the conversion of turn-motif and cyclic peptides into drug candidates, noting that the structural rigidity introduced by cyclisation is a prerequisite for oral or systemic stability rather than an optional refinement. Co-crystal validation at this sub-stage confirms that the bridge does not sterically clash with receptor residues and that the constrained helix aligns with the binding groove geometry established in Phase 1.

Backbone N-Methylation

Replacing a backbone NH with N-methyl reduces amide bond susceptibility to proteolytic cleavage and modulates membrane permeability by eliminating a hydrogen-bond donor. Structural re-determination after N-methylation is non-negotiable: the methyl group projects into the binding site and can displace water molecules that mediate receptor contacts, an effect invisible without a new co-crystal dataset.

Lipid Conjugation and Half-Life Extension

Lipid conjugation extends plasma half-life by promoting reversible albumin binding, slowing renal clearance, and reducing receptor-mediated degradation. The GLP-1 series provides the clearest phase-annotated illustration of how this strategy evolved structurally:

• Exenatide, a 39-residue exendin-4 analogue with no lipid conjugation, requiring twice-daily injection. Its helical scaffold was geometrically compatible with the GLP-1R extracellular domain groove but offered no intrinsic half-life advantage.
• Liraglutide, a C18 fatty acid attached via a glutamic acid spacer at Lys26; once-daily dosing achieved through albumin binding. Co-crystal data confirmed that the acyl chain projects away from the receptor interface, leaving the helical binding epitope intact.
• Semaglutide, a C18 diacid conjugated through a longer hydrophilic linker, with two additional backbone modifications that reduce DPP-4 cleavage at position 2; once-weekly dosing. Cryo-EM structures of semaglutide bound to GLP-1R (deposited 2021–2023) show the receptor transmembrane bundle in an active conformation essentially identical to that seen with shorter analogues, validating that the extended linker does not remodel the orthosteric site. Hornsby et al. (2026) note that peptides and proteins must be chemically and structurally optimised to achieve stability for effective drug function. Further information on the high-dose semaglutide formulation approved by the EMA in 2025–2026 is available at the JCSG semaglutide resource.
• Tirzepatide, a dual GIP/GLP-1 agonist whose sequence was redesigned from first principles using cryo-EM structures of both GLP-1R and GIPR in active-state complexes with surrogate agonists. The structural redesign identified a single peptide scaffold capable of adopting receptor-compatible helical geometries at both binding sites, a conclusion that single-receptor docking could not have reached. Tirzepatide holds EMA marketing authorisation for type 2 diabetes and weight management; structural and clinical context is collated at the JCSG tirzepatide page.

Across all four compounds, the co-crystal or cryo-EM dataset obtained after each chemical modification was the gate that determined whether the candidate advanced. The JCSG pipeline methodology describes how iterative structure determination is embedded in a documented technology workflow rather than treated as an ad hoc confirmatory step, and the JCSG Structure Gallery provides deposited coordinates across receptor families that serve as reference geometries when calibrating whether a modified peptide retains its intended binding mode.

The GLP-1 Series: A Phase-Annotated Case Study

The GLP-1 agonist series is the clearest available illustration of how successive structural datasets, each obtained after a discrete chemical modification, drove each candidate from a natural peptide scaffold to a multi-receptor dual agonist with EMA marketing authorisation.

The table below maps each molecule to the structural insight that determined its design trajectory. Regulatory status reflects EMA records current to mid-2026; see the researcher warning beneath the table before citing approval dates in any submission or publication.

Molecule	Key Structural Modification	Structural Data Type	Primary Optimisation Goal	EMA Approval (approx.)
Exenatide (Byetta)	Native exendin-4 scaffold; no synthetic modification to core sequence	X-ray co-crystal of GLP-1R extracellular domain (ECD) with peptide N-terminus	Confirm receptor-binding geometry of a non-mammalian GLP-1 analogue; establish ECD contact map as design template	2006 (EU)
Liraglutide (Victoza/Saxenda)	C18 fatty acid attached via glutamate-linker at Lys²⁶; Arg³⁴Lys substitution	X-ray co-crystal of albumin–fatty acid complex; NMR of peptide in solution	Extend half-life through reversible albumin binding; retain helical receptor-binding conformation in solution	2009 (T2D); 2015 (obesity)
Semaglutide (Ozempic/Wegovy)	C18 diacid linker at Lys²⁶; Aib substitution at position 8 for DPP-4 resistance; two mini-PEG spacers	Cryo-EM of full-length GLP-1R in active-state complex; solution NMR for linker geometry	Maximise albumin affinity and proteolytic stability simultaneously; weekly dosing; high-dose oral formulation approved by EMA in early 2026	2018 (T2D); 2021 (obesity); 2026 (high-dose oral)
Tirzepatide (Mounjaro)	Single 39-residue scaffold redesigned to present GIP-compatible N-terminus and GLP-1-compatible mid-helix; C18 diacid at Lys²⁰	Cryo-EM structures of GLP-1R and GIPR active-state complexes used in parallel; no single-receptor docking could resolve dual geometry	Achieve equipotent dual agonism at GIP-R and GLP-1R from one peptide; EMA-authorised for T2D and weight management	2022 (T2D); 2023 (obesity)

For researchers working on next-generation triple agonists, the JCSG tirsema resource collates structural and formulation context for high-dose GIP/GLP-1 combination approaches, and the JCSG retatrutide page covers the glucagon-receptor arm that distinguishes triple from dual agonism. The JCSG Structure Gallery provides deposited reference coordinates across receptor families against which modified peptide binding modes can be calibrated at each design iteration.

Researcher warning, evolving regulatory status. GLP-1 agonist approvals and indications are changing rapidly: the EMA granted a positive opinion for high-dose oral semaglutide in 2024 and European Commission approval followed in early 2026. Tirzepatide's EPAR continues to be updated. Verify all approval dates, indications, and formulation authorisations against current EMA EPAR records before citing this table in any publication, regulatory submission, or clinical guidance document.

Phase 4, Preclinical Validation: Structure Informs Off-Target Risk

Validated leads entering cellular and animal models do not leave structural biology behind: deposited coordinates continue to drive selectivity engineering, ADMET risk assessment, and conformation-consistency checks throughout preclinical work.

Selectivity Engineering Through Binding-Site Topology Comparison

Off-target activity at a paralogue receptor is the most tractable structural risk to address before in vivo studies begin. For GPCR-targeting peptides, the challenge is acute: GLP-1R, GIPR, and the glucagon receptor share a class B GPCR fold with overlapping extracellular domain topologies, meaning a peptide optimised against one receptor can cross-react with either paralogue. Structural comparison of the three binding sites (using PDB-wide searches or purpose-built tools such as the JCSG Protein Structure Comparison Analyser (PSCA)) quantifies the degree of pocket similarity before any animal experiment is run. In tirzepatide development, this comparison was not a precaution but a design objective: the dual GIP/GLP-1 scaffold was deliberately engineered to exploit shared topology at both receptors while preserving selectivity away from the glucagon receptor, a distinction that required parallel cryo-EM datasets of all three active-state complexes to resolve (Iglesias et al., 2024).

ADMET Profiling Informed by Structural Data

Proteolytic stability is the dominant ADMET liability for peptide leads. Structural data identifies exposed backbone segments and flexible termini susceptible to endopeptidase cleavage, allowing medicinal chemists to prioritise α-methylation, N-methylation, or cyclisation at vulnerable positions before committing to in vivo pharmacokinetic studies. Membrane permeability modelling for oral candidates draws on the same coordinate sets: solvent-accessible surface area calculations derived from deposited structures feed directly into passive permeability predictions. Solid-phase peptide synthesis (SPPS) is the standard manufacturing route for preclinical quantities at this stage, providing milligram-to-gram amounts of modified leads with defined stereochemistry (Fetse et al., 2023).

Revisiting Multi-Conformation Datasets

The multi-conformation datasets assembled during Phase 1 target characterisation are revisited at this point to verify that the optimised lead still binds the physiologically relevant receptor conformation rather than a crystallographic artefact. If ensemble modelling from Phase 1 identified an inactive-state pocket that is absent in the active-state structure, any lead that docks preferentially to that pocket requires re-evaluation before animal dosing. The JCSG Structure Gallery and the JCSG pipeline methodology document how multi-conformation reference sets are curated and made available for exactly this cross-phase consistency check, providing the deposited coordinate infrastructure that makes the comparison reproducible across research groups.

Structural information in therapeutic peptide development is a documented de-risking mechanism at the preclinical stage, reducing the probability that selectivity failures or conformational mismatches surface for the first time in costly in vivo studies (Iglesias et al., 2024).

Phase 5, Clinical Translation: Cryo-EM and Near-Physiological Structures

Structural biology does not stop contributing when a peptide therapeutic enters clinical trials; cryo-EM data acquired during and after Phase 3 now routinely informs biomarker selection, mechanistic labelling, and indication expansion for GPCR-targeting peptides.

Cryo-EM as a Clinical-Stage Instrument

Cryo-EM structures of therapeutic peptides bound to their receptors under near-physiological conditions have become standard evidence for class B1 GPCRs. For the GLP-1 series, cryo-EM complexes of semaglutide bound to GLP-1R and of tirzepatide bound to both GLP-1R and GIPR were published in the 2021–2023 period, resolving the active-state receptor conformation, G-protein coupling geometry, and the precise peptide binding register at sub-3 Å resolution. These datasets are qualitatively distinct from earlier crystal structures because they capture the receptor in a lipid-bilayer-mimicking nanodisc or detergent environment, removing the lattice-packing constraints that can distort transmembrane helix geometry. The mechanistic depth of protein–peptide interactions at this resolution is documented in the JCSG resource on how protein–peptide interactions shape cellular function, which situates receptor-level contacts within broader signalling consequences.

Structural Data in Regulatory Submissions

Mechanistic understanding of binding mode can inform the benefit–risk assessment presented to regulators. The EMA's CHMP issued a positive opinion in May 2024 recommending approval of both an oral semaglutide tablet and a high-dose injectable pen of Wegovy for weight management, with the European Commission granting approval in early 2026. Structural evidence supporting receptor selectivity and the absence of off-target engagement at related receptors contributes to the mechanistic sections of a marketing authorisation application, even though it does not substitute for clinical safety data. Wang et al. (2022) reviewed how structural characterisation of peptide–receptor complexes now features explicitly in the mechanistic rationale sections of regulatory dossiers for peptide therapeutics.

Structure-Guided Label Expansion

Where cryo-EM reveals a second binding mode or an allosteric site not apparent in the primary active-state structure, that finding can support a new indication filing. Tirzepatide's dual GLP-1R and GIPR agonism, confirmed structurally, underpins its EMA-approved indications for both type 2 diabetes and weight management. The JCSG Structure Gallery and JCSG pipeline methodology document the multi-conformation coordinate infrastructure that makes cross-receptor comparisons of this kind reproducible across independent research groups.

Disclaimer: All pipeline discussion in this article is scientific and educational in nature. Nothing here constitutes medical, clinical, or regulatory guidance, and readers should consult current EMA, MHRA, or equivalent regulatory documentation for authoritative product information.

What Structural Genomics Adds That Single-Structure Studies Cannot

Structural genomics programmes generate systematic, multi-condition datasets across entire protein families, and that breadth is precisely what single-structure studies cannot replicate. A lone crystal structure captures one conformational snapshot under one set of crystallisation conditions; a structural-genomics campaign captures the same target in multiple crystal forms, pH conditions, and ligand-bound states, producing an ensemble that reflects the genuine conformational range a drug candidate must navigate in solution.

Multi-Conformation Capture

When a target is crystallised under varied conditions (different precipitants, temperatures, and co-solutes), the resulting ensemble reveals hinge motions, loop rearrangements, and pocket breathing that a single structure obscures. For peptide drug design, this matters acutely: a binding groove that appears closed in one crystal form may be fully accessible in another, and a lead optimised against only the closed form will fail to exploit that flexibility. Iglesias et al. (2024) found that integrating multiple structural states, rather than relying on a single representative coordinate set, substantially improves the accuracy of binding-site characterisation for peptide–receptor systems. The JCSG Structure Gallery illustrates this principle directly, documenting how systematic deposition across varied crystallisation conditions builds the conformational coverage that informs downstream lead optimisation.

Domain Boundary Precision

High-throughput construct screening, a defining feature of structural-genomics pipelines, identifies the minimal stable domain that yields diffraction-quality crystals. This is not a trivial benefit: constructs that are too long introduce flexible termini that disorder the lattice, while constructs that are too short truncate functionally relevant secondary structure. Precise domain boundaries reduce false negatives in binding assays by ensuring the recombinant protein presents the same folded surface as the native target. The JCSG pipeline methodology documents the construct-screening workflow that underpins this precision.

Homologue Coverage and Selectivity Mapping

Structural-genomics programmes deposit structures of paralogue and orthologue proteins alongside the primary target. For a peptide therapeutic targeting a class B GPCR such as GLP-1R, having deposited structures of related receptors (GIPR, glucagon receptor, GLP-2R) allows medicinal chemists to map selectivity determinants directly from coordinate comparisons rather than inferring them from sequence alignment alone. This cross-receptor structural coverage is what enabled the mechanistic rationalisation of tirzepatide's dual GLP-1R/GIPR agonism. Further mechanistic context on how deposited coordinates support peptide–receptor interaction analysis is available at the JCSG resource on understanding protein–peptide interactions.

Open-Access Deposition

Structural genomics is defined by its deposition mandate: all coordinates enter the PDB under standard wwPDB open-access terms, meaning any research group can build on the data without repeating expensive structure determination. This open model accelerates protein structure to peptide therapeutic development across the community, not just within the originating laboratory. Researchers using JCSG-deposited multi-conformation datasets should verify current accession status and coordinate completeness against live PDB records and JCSG resources, as database contents and any residual hold periods should always be confirmed against current wwPDB policy rather than assumed from prior literature.

Key Challenges and Open Questions in 2026

Structure-guided peptide therapeutic development faces five substantive barriers that no single methodological advance has yet resolved: oral delivery, immunogenicity prediction, manufacturing cost, the limits of AI-predicted structures, and regulatory ambiguity around hybrid and stapled peptide formats.

Oral Bioavailability

The majority of approved peptide therapeutics remain injectable because intestinal proteolysis and poor membrane permeability combine to suppress oral bioavailability for molecules above roughly 500 Da. Cyclisation strategies and permeation enhancers such as SNAC (used in oral semaglutide) have demonstrated proof of concept, but neither approach is yet routine for peptides exceeding 30 residues. Hayward et al. (2024) document cyclic peptide conversion strategies that exploit deposited backbone conformations to constrain ring geometry, yet translating those conformations into reliably absorbed oral candidates remains a case-by-case challenge rather than a generalised workflow.

Immunogenicity Prediction

Deposited coordinates reveal surface-exposed epitopes and can guide the removal of T-cell receptor contact residues, but structural data alone cannot predict the immunogenic response in a heterogeneous patient population. Sequence-based tools and MHC-binding algorithms supplement structural analysis, though the integration of multi-conformation datasets into immunogenicity workflows is still an active research area rather than an established regulatory expectation.

Manufacturing Scale-Up

Solid-phase peptide synthesis (SPPS) remains the dominant production route for peptides below approximately 50 residues, but commercial-scale SPPS is reagent-intensive and generates substantial solvent waste. Hornsby et al. (2026) identify manufacturing cost and synthetic accessibility as constraints that feed back into lead optimisation decisions: structural modifications that improve potency or stability must be weighed against the synthetic steps they introduce.

AI-Assisted Structure Prediction

AlphaFold 3, released in 2024, extended earlier protein-only prediction to protein–peptide and protein–ligand complexes, offering a faster route to structural hypotheses at Phase 1 and Phase 2 of the pipeline. Its documented limitations are significant, however: reliability drops for large conformational changes on binding, flexible or disordered binding regions are handled poorly, and quantitative binding energetics are outside its scope. As of 2026, no EMA submission cites AlphaFold 3 outputs as pivotal evidence; predicted structures remain supportive rather than primary, and experimental deposition through programmes documented in the JCSG Structure Gallery continues to provide the ground-truth coordinates that AI models are trained and benchmarked against.

Regulatory Ambiguity for Hybrid and Stapled Peptides

Stapled peptides and peptide-protein conjugates occupy a classification grey zone between small molecules and biologics. EMA guidance on which regulatory pathway applies (and which manufacturing and immunogenicity data packages are required) is still evolving. Researchers should treat any pipeline success-rate statistics predating 2023 as potentially stale, given that the approval environment for these hybrid modalities has shifted materially since the first stapled peptide candidates entered clinical evaluation. The JCSG pipeline methodology illustrates how systematic structural coverage of target families can at least reduce mechanistic uncertainty at the point of regulatory submission, even where classification questions remain open.

Summary: The Five Phases at a Glance

Each phase of the protein structure to peptide therapeutic development pipeline draws on deposited structural-genomics coordinates as an active decision-making input, not a retrospective record.

Phase	Primary structural input	Key decision enabled	GLP-1 series milestone
1 · Target Characterisation	Apo-receptor crystal structures; multi-conformation ensemble datasets	Confirm druggable binding site; define conformational states	GLP-1R extracellular domain structures establish ECD and TMD as tractable targets
2 · Lead Identification	Peptide–receptor co-crystal and cryo-EM complexes	Select native peptide scaffold; map critical contact residues	GLP-1(7–36) binding mode at GLP-1R defines the α-helical pharmacophore
3 · Lead Optimisation	High-resolution cryo-EM of analogue–receptor complexes	Guide residue substitution, fatty-acid conjugation, and stapling	Semaglutide cryo-EM structures (deposited 2021–2023) rationalise C18 diacid linker positioning
4 · Preclinical Validation	Comparative structural datasets across species orthologues	Confirm cross-species binding geometry; support toxicology species selection	Rodent and primate GLP-1R structural homology validates rat and cynomolgus models
5 · Clinical Translation	Cryo-EM of approved-drug complexes; dual-agonist structures	Inform next-generation candidates; support regulatory mechanistic packages	Tirzepatide GIP-R/GLP-1R dual-agonist structures (deposited 2022–2024) underpin EMA approval dossier

Deposited coordinates at each phase reduce mechanistic uncertainty before costly in vivo experiments begin; the GLP-1 series demonstrates this concretely, with cryo-EM datasets driving decisions from initial scaffold selection through to the EMA approvals of semaglutide and tirzepatide for type 2 diabetes and weight management. Semaglutide's 2024 CHMP positive opinion extending Wegovy's authorisation to a high-dose oral formulation was supported by receptor-level structural understanding established years earlier in the PDB, illustrating how a single deposition programme can sustain multiple regulatory submissions across a decade.

Next Steps

Researchers seeking to apply this framework to their own targets can explore the JCSG pipeline methodology, review target coverage through the Protein Structure Coverage Analyser, and consult the mechanistic context provided in JCSG's account of how protein–peptide interactions shape cellular function. The full suite of deposited coordinates and associated data is accessible via JCSG Resources.

To discuss structural coverage of your target family or explore collaboration on structure-guided peptide programmes, contact the JCSG team directly through the resources page.

Related JCSG research

• How Protein–Peptide Interactions Shape Cellular Function — the SLiM–domain logic that underpins these therapeutics.
• Understanding Protein–Peptide Interactions — affinity, selectivity, and how binding is measured.
• Protein–Peptide Interactions in Structural Genomics — mapping interaction interfaces through high-throughput structure determination.
• Structural Genomics in Peptide Drug Discovery and the role of structural genomics in advancing candidates.
• Explore the JCSG Structure Gallery, the gene-to-structure pipeline, and JCSG resources & tools.

References

Key studies and sources referenced in this article:

1. Wang et al. (2022) — reviewed how structural characterisation of peptide–receptor complexes now features explicitly in the mechanistic-rationale sections of regulatory dossiers for peptide therapeutics.
2. Fetse et al. (2023) — demonstrated that peptoids retain structural motifs sufficient to engage peptide-binding pockets, making them useful leads where proteolytic stability is a concern.
3. Hayward et al. (RSC, 2024) — documented the conversion of turn-motif and cyclic peptides into drug candidates, with cyclisation-induced rigidity as a prerequisite for oral and systemic stability.
4. Iglesias et al. (2024) — resolved active-state receptor complexes underpinning the GIP/GLP-1 selectivity engineering behind tirzepatide.
5. Hornsby et al. (2026) — noted that peptides and proteins must be chemically and structurally optimised to achieve the stability required for effective drug function.